PhD. Thesis defence Slides

1. 104th PhD thesis defense at PPGEE/UFPA
Cross-layer Optimizations for Multimedia
Distribution over WMSNs and FANETs with QoE
Support
Denis Lima do Rosário
Advisors: Eduardo Coelho Cerqueira
Torsten Braun

2. Outline
´ Introduction
´ Multi-hop and multi-path hierarchical routing protocol
for Efficient VIdeo transmission (MEVI)
´ Cross-layer Link quality and Geographical-aware
beaconless Opportunistic routing protocol (XLinGO)
´ Conclusion and Future Work
2

3. Introduction – Wireless Multimedia
Sensor Network (WMSN) Scenario
3

4. Introduction – Flying Ad-hoc Network
(FANET) Scenario
Systems Mobile
UsersBase
Station
Control
Center
4

5. Requirements for a Real-time
Multimedia Distribution I
´ Scalability
´ Routing protocols must provide an efficient multi-hop communication
between any pair of source and destination nodes, in order to deliver the
collected data from any part of the monitored area.
´ The multi-hop communication must be scalable without requiring user
intervention, and also independently of the number of nodes or field size
´ Efficient Buffer Control
´ Multimedia dissemination usually involves a set of nodes transmitting
multiple video flows simultaneously.
´ Leading to a higher degree of network congestion, buffer overflow, and
packet loss ratio, which reduces the quality level of the delivered video
flows.
´ The routing protocol must prevent the selection of forwarding nodes with
heavy traffic load.
5

6. Requirements for a Real-time
Multimedia Distribution II
´ Robustness
´ The network nodes must cope with dynamic topologies caused by
node failure or mobility, and wireless channel changes
´ The multimedia dissemination continue reliable and robust despite
dynamic topologies.
´ Energy-efficiency
´ Energy consumption is also prime concern in WMSNs and
FANETs
´ Both networks consist of battery-powered nodes with limited
energy resources.
´ The development of energy-efficient communication protocols is
one of the main goals to increase network lifetime.
6

7. Requirements for a Real-time
Multimedia Distribution III
´ Quality of Experience (QoE) Requirements
´ Solutions involving multimedia transmissions must evaluate the
video content from the user’s perspective and not only from the
network’s perspective.
´ Over the last decade the focus has shifted away from pure
network point-of-view assessment (QoS metrics) to a more
human-centric approach (QoE metrics) and user-awareness.
´ QoS schemes alone are not enough to assess the quality level of
multimedia applications, because they fail to capture subjective
aspects of video content related to human experience and
subjectivity
7

8. Requirements for a Real-time
Multimedia Distribution IV
´ Unreliable Nature of Wireless Channels
´ Link quality estimation is a fundamental building block in the
design of routing protocols for WMSN and multimedia
FANET scenarios
´ A reliable routing protocol must consider the link quality as a
metric to select reliable quality routes for multimedia
dissemination.
8

9. Main Research Question and
Contributions
´ The main research contributions of this thesis addresses by the research
question of how to provide real-time multimedia distribution with high
energy-efficiency, reliability, robustness, scalability, and QoE support
over wireless ad-hoc networks.
´ Multi-hop and multi-path hierarchical routing protocol for Efficient VIdeo
transmission for static WMSN scenarios (MEVI)
´ Cross-layer Link quality and Geographical-aware beaconless OR protocol for
multimedia FANET scenarios (XLinGO)
´ QoE-aware Packet-Level Redundancy Mechanism (QoE-aware
redundancy)
´ Mobile MultiMedia Wireless Sensor Network (M3WSN) OMNeT++
framework
9

10. Thesis Contributions
Source Node Destination NodeIntermediate Node
Physical
Application
Transport
MAC
Routing
Wireless ad-hoc network
Physical
MAC
Routing
Physical
Application
Transport
MAC
RoutingXLinGO or MEVIXLinGO or MEVI
QoE-aware
redundancy
10

11. Related Work
´ The existing hierarchical routing protocols have the
following drawbacks:
´ Include a higher overhead to create cluster.
´ Lack of efficient multi-hop and multi-path communication.
´ Lack of reliable cross-layer approach to select routes based on
information about.
´ End-to-end link quality estimation;
´ Energy;
´ Number of Hops.
´ The existing beaconless OR protocols have the
following drawbacks:
´ Do not consider multiple metrics for forwarding decisions.
´ Do not quickly detecting and responding to topology changes.
11

12. Outline
´ Introduction
´ MEVI
´ Design and Operation Principles
´ Simulation Environment and Metrics
´ Evaluation
´ XLinGO
´ Conclusion and Future Work
12

13. Multi-hop and multi-path hierarchical routing protocol
for Efficient VIdeo transmission (MEVI)
´ MEVI is designed to work in a WMSN application with a
fixed network infrastructure to accurately monitor physical
scalar measurements, and also collect multimedia data in
the case of an event occurrence.
´ MEVI relies on:
´ Hierarchical architecture;
´ Heterogeneous nodes;
´ MEVI considers two phases for data transmission:
´ Intra-cluster communication;
´ Inter-cluster communication.
13

14. MEVI: Intra-Cluster Communication
´ Network nodes create
clusters.
´ Cluster members sending
the sensed value in a
specific slot to their
Cluster Head (CH).
´ Clusters are created with
low signaling overhead,
since nodes only send:
´ Beacons;
´ Data.
14

15. MEVI: Inter-Cluster Communication
MEVI exploits a reactive scheme
to find on-demand multiple paths.
• Route request; and
• Route reply messages.
Each possible path has a Path
Quality (PQ) associated.
• E n d - t o - e n d l i n k q u a l i t y
estimation;
• Energy;
• Number of Hops.
15

16. MEVI: Inter-Cluster Communication
CHs must send the aggregate
packet to the DN.
16

17. MEVI: Inter-Cluster Communication
DN must analyze the received
data by means of existing models
or methods.
It requests a video sequence from
a given CH, as soon as it detects
an event occurrence.
17

18. MEVI: Inter-Cluster Communication
M E V I s c h e d u l e s p a c k e t
transmissions via multiple paths to
provide robustness and load
balancing.
MEVI schedules the transmission
of priority frames via the best path
• I-frame; and
• first P-frames.
18

19. Outline
´ Introduction
´ MEVI
´ Design and Operation Principles
´ Simulation Environment and Metrics
´ Evaluation
´ XLinGO
´ Conclusion and Future Work
19

20. Mobile MultiMedia Wireless Sensor
Network OMNeT++ framework (M3WSN)
´ M3WSN framework relies on Castalia architecture to
provide new functionalities.
´ It implements full support to:
´ deliver, control, and evaluating
real video sequences.
´ scenarios composed of fixed
and mobile nodes, as well as
moving object.
20

21. Evaluation Metrics
´ Energy-Efficiency Evaluation
´ Network lifetime
´ time spent until 10% of the network nodes remain alive; or
´ the moment of the first node run out of energy resources.
´ Objective and Subjective QoE Evaluation
´ Structural SIMilarity (SSIM) measures the structural
distortion of the video, and attempts to obtain a better
correlation with the user’s subjective impression.
´ Transmitted frame
21

22. Outline
´ Introduction
´ MEVI
´ Design and Operation Principles
´ Simulation Environment and Metrics
´ Evaluation
´ XLinGO
´ Conclusion and Future Work
22

23. Evaluation Goal
´ Simulation experiments aim to show the scalability,
reliability, and energy-efficiency of MEVI for
transmitting multimedia content compared to existing
hierarchical routing protocols for a static WMSN
scenarios.
23

24. Hierarchical Routing Protocols under Evaluation
´ MEVI
´ Nodes create clusters with low overhead.
´ Nodes classify routes based on multiple metrics.
´ It considers a multi-path video transmission
´ LEACH
´ Nodes have homogeneous capabilities and transmit beacon, join, and
schedule messages for cluster formation.
´ LEACH fixed CHs
´ Nodes have heterogeneous capabilities, i.e., it considers fixed CHs.
´ It follows the traditional LEACH implementation for cluster formation.
´ PEMuR:
´ nodes periodically create clusters in a centralized way by transmitting
beacon, schedule, advertisement, identifier, and also join messages.
´ Nodes classify routes based only on remaining energy.
24

25. Network Lifetime
LEACH fixed CHs
25

26. Video Quality
MEVIPEMuR
LEACH LEACH fixed CHs
26

27. Original MEVI
LEACH PEMuR
27
Frame of transmitted video27

28. MEVI in Summary
´ It considers a cluster formation with low overhead.
´ provides efficient and reliable intra-cluster communication.
´ It triggers the multimedia transmission according to the
sensed value.
´ avoids the transmission of unnecessary video content and saving
scarce network resources, such as energy and bandwidth.
´ It finds a subset of reliable CHs to establish multiple paths.
´ the subset of optimal CHs must provide high packet delivery rate.
´ It schedules multimedia transmission via multi-paths
according to the frame relevance
´ provides load balancing, robustness, and QoE-awareness.
28

29. Outline
´ Introduction
´ MEVI
´ XLinGO
´ Design and Operation Principles
´ Evaluation
´ Conclusion and Future Work
29

30. Cross-layer Link quality and Geographical-
aware beaconless OR protocol (XLinGO)
´ XLinGO is designed to work in multimedia FANET application to
explore, sense, and also send multimedia data from the hazardous
area as soon as the fixed network infrastructure is not available.
´ The main goal is to deliver multimedia content with a minimum video
quality level from user’s perspective.
´ XlinGO relies on a beaconless OR approach
´ Takes into account multiple metrics for forwarding decision.
´ Considers a recovery mechanism for route reconstruction.
´ Includes two operational modes:
´ Contention-based forwarding mode;
´ Persistent route mode.
30

31. Opportunistic Routing - coordination
method
´ Beacon-based OR
´ Selects and prioritizes a set of candidate nodes by transmitting beacon
messages before packet transmission.
´ Creates and orders a relay candidate list prior to packet transmission
according to a certain criteria, such as expected transmission count.
´ Determines the candidate list before sending packets, and may not
reflect the real situation at the moment of packet transmission.
´ Beaconless OR
´ Forwarding decisions are performed by the receiver of a packet.
´ Adds a small Dynamic Forwarding Delay (DFD) value before packet
transmission.
´ Candidate node with best conditions compute the shortest DFD value,
and thus such node transmits the packet faster.
31

32. XLinGO: Contention-based forwarding
mode
32

33. XLinGO: Contention-based forwarding
mode
1. Source broadcasts the video packet
33

34. XLinGO: Contention-based forwarding
mode
1. Source broadcasts the video packet
2. Nodes analyze the forwarding area
34

35. XLinGO: Contention-based forwarding
mode
1. Source broadcasts the video packet
2. Nodes analyze the forwarding area
3. Nodes inside PPA compute the DFD based on geographical information, queue size, energy,
and link quality.
35

36. XLinGO: Contention-based forwarding
mode
1. Source broadcasts the video packet
2. Nodes analyze the forwarding area
3. Nodes inside PPA compute the DFD based on geographical information, queue size,
energy, and link quality
4. Node that generates the smallest DFD forwards the packet faster
36

37. XLinGO: Contention-based forwarding
mode
37

38. XLinGO: Persistent forwarding mode
SN
Fi
Data packets
…
1. SN sends n video packet to Fi
2. Fi computes the average of link quality and PDR for the last n received packets
3. Fi sends a reply message to SN and piggyback link quality and PDR
4. SN evaluates if it continues transmitting packets to Fi or if it returns to the Contention-based
forwarding mode
38

39. Outline
´ Introduction
´ MEVI
´ XLinGO
´ Design and Operation Principles
´ Evaluation
´ Conclusion and Future Work
39

40. Evaluation Metrics and Goal
´ Simulation experiments aim
´ to show the performance of the recovery mechanism to quickly
detect and react to topology changes.
´ to evaluate different beaconless OR protocol in a scenario with
node mobility and simultaneous multiple video flows
dissemination.
´ Objective and Subjective QoE Evaluation
´ Structural SIMilarity (SSIM) measures the structural distortion of
the video, and attempts to obtain a better correlation with the
user’s subjective impression.
´ Transmitted video.
40

41. Beaconless OR Protocols under Evaluation
´ Original video
´ represents an errorless video transmission.
´ XLinGO
´ does not have node failures and recovery
mechanism;.
´ XLinGO – Failure
´ relies on periodic route reconstruction.
´ experiences node failures.
´ XLinGO – Recovery
´ considers the recovery mechanism and
experiences node failures.
41

42. Impact of the mechanism to
recover from route failures
Interval for
route
reconstruction
Timeout to
detect route
failure
42

43. Beaconless OR Protocols under Evaluation
´ XLinGO
´ combines queue length, link quality, geographical location, and residual energy to compute the DFD.
´ relies on a recovery mechanism to react faster to route failure situations.
´ LinGO
´ combines link quality, geographical information, and energy to compute the DFD.
´ relies on periodic route reconstruction.
´ BLR
´ considers only geographical information to compute the DFD.
´ relies on periodic route reconstruction.
´ BOSS
´ relies only on geographical information to compute the DFD.
´ consider a tree-way handshake.
´ relies on periodic route reconstruction.
´ MRR
´ compute the DFD function based on geographical information, RSSI, and energy.
´ relies on a periodic route reconstruction.
43

44. Simulation Description
´ Field size of 100x100
´ Deployed 40 nodes
´ 1 – Destination node (static)
´ 2 – Mobile source nodes
´ 38 – Mobile possible forwarder nodes
´ Transmitted videos
´ Container, Hall, and Highway (YUV library)
´ UAV1 and UAV2 (downloaded from Youtube)
44

45. Impact of Node Mobility45

46. Transmitted video
Video Transmitted via XLinGO Video Transmitted via BLR
More videos at: https://plus.google.com/b/102508553201652207043/102508553201652207043/posts
46

47. XLinGO in Summary
´ It finds a subset of reliable forwarders to establish a persistent
route
´ the subset of optimal forwarders provides high packet delivery rate,
enabling video delivery with high video quality level from the user’s
experience.
´ It prevents the selection of a forwarders with heavy traffic load
or low residual energy.
´ providing load balancing and energy-efficiency, as well as reduce queue
congestions, packet loss, delay, and jitter.
´ It determines whether one of the forwarders from a given
persistent route might be no longer available or still reliable to
forward packets.
´ enabling a smoother operation in mobile networks.
´ providing robust and reliable multimedia transmission.
47

48. Outline
´ Introduction
´ MEVI
´ XLinGO
´ Conclusion and Future Work
48

49. Conclusion
´ We addressed the main research question of this thesis
´ how to provide real-time multimedia distribution with high energy-efficiency, reliability,
robustness, scalability, and QoE support over wireless ad-hoc networks.
´ We developed:
´ Multi-hop and multi-path hierarchical routing protocol for Efficient VIdeo transmission
for static WMSN scenarios (MEVI)
´ Cross-layer Link quality and Geographical-aware beaconless OR protocol for multimedia
FANET scenarios (XLinGO)
´ QoE-aware Packet-Level Redundancy Mechanism (QoE-aware redundancy)
´ Mobile MultiMedia Wireless Sensor Network (M3WSN) OMNeT++ framework
´ We consider two classes of applications.
´ WMSN application that relies on fixed network infrastructure to accurately monitor
physical scalar measurements, and also collect multimedia data in the case of an event
occurrence.
´ Multimedia FANET application to explore, sense, and also send multimedia data from
the hazardous area as soon as the fixed network infrastructure is not available.
49

50. Conclusion
Source Node Destination NodeIntermediate Node
Physical
Application
Transport
MAC
Routing
Wireless ad-hoc network
Physical
MAC
Routing
Physical
Application
Transport
MAC
Routing
MEVI and
XLinGO
MEVI and
XLinGO
QoE-aware
redundancy
´ We address several problem domains with contributions on
different layers of the communication stack.
50

51. Conclusion
´ The performance and behavior of our contribution
for multimedia distribution was evaluated by means of
simulation experiments.
´ Based on simulation results, we conclude that our
cross-layer optimizations for multimedia distribution
over WMSN and multimedia FANETs achieved results
that filled the goal of our initial research question.
*All developed modules will be available at http://cds.unibe.ch/research/M3WSN/
51

52. Future works
´ Perform Testbed experiments and long-term real-
world deployment to confirm the simulation results.
´ TARWIS
´ XLinGO must consider the link expiration time estimation,
position prediction, and moving direction in a 3D plane for
routing decision.
´ avoiding the selection of a forwarder node that is moving in opposite
direction in a 3D plane than its previous hop, which avoids loops and
prevents suboptimal routing.
52

53. Future works
´ Several issues need to be studied and understood in order to
extend MEVI and XLinGO by taking into consideration, such
as human-centric schemes, video characteristics, and context-
awareness for decision making.
´ online QoE assessment and user experience can be measured and
integrated into MEVI and XLinGO protocols, in order to support
routing decisions to improve the user satisfaction on watching real-time
video flows.
53

54. List of Publications in Summary
´ 20 publications in national or international refereed journals,
conferences, or workshops
´ 2 - Qualis A1 (2 journals)
´ 2 - Qualis A2 (1 journal and 1 conference)
´ 1 - Qualis B1 (1 conference)
´ 1 - Qualis B2 (1 conference)
´ 2 - Qualis B3 (2 conferences)
´ 2 - Qualis B4 (2 conferences)
´ 1 - Qualis B5 (1 journal)
´ 2 - Book Chapters
´ 7 - Qualis not Available
*Following the ENG IV Qualis for journals papers, and CC Qualis for conference papers
54